Analyzing Meshes ¶
Very often there will be specific regions of a set of results that are the most of interest, and in those cases it can be helpful to pull out a particular region's data for visualization in a clearer way. The
LineCrossSection
and
SurfaceCrossSection
classes allow taking cross-sections of a mesh (or some other rasterizable object) along poly-lines drawn using the
PolyDraw
tool, letting you make an
x
/
z
or
x
/
z
/
t
plot from data laid out in
x
/
y
/
z
or
x
/
y
/
z
/
t
(respectively). These classes rasterize the data at a fixed resolution, and then sample the data at that resolution given the underlying polylines. The
resolution
may be defined in the units of the underlying coordinate system.
import param
import numpy as np
import datashader as ds
import holoviews as hv
import geoviews as gv
import cartopy.crs as ccrs
from colorcet import cm_n
from earthsim.analysis import LineCrossSection, SurfaceCrossSection
from earthsim.io import read_3dm_mesh, read_mesh2d
hv.extension('bokeh')
%output holomap='scrubber'
%opts Image RGB [width=500 height=400 colorbar=True] Curve {+framewise} [height=400 width=400] VLine (color='black')
%opts Path (line_width=3 color='black') NdOverlay [legend_limit=0] Image {+framewise} [height=400 width=500]
Example 1: A static TriMesh ¶
As a first simple example, we will pass a static TriMesh to the
LineCrossSection
and display it. We'll pass a couple of sample polylines to start with, and if a live Python process is available we will be able to see the depth along the paths appear on the right, for those paths and any more that are subsequently drawn by the user.
filename = '../data/Chesapeake_and_Delaware_Bays.3dm'
cb_paths = [[(-8523594., 4588993.), (-8476533., 4578872.),
(-8449931., 4562126.), (-8435409., 4539747.)],
[(-8477265., 4544109.), (-8469725., 4430387.)]]
tris, verts = read_3dm_mesh(filename)
points = gv.operation.project_points(gv.Points(verts, vdims=['z']))
trimesh = hv.TriMesh((tris, points))
sector1 = LineCrossSection(trimesh, cb_paths)
sector1.view()
Each individual polyline (thick black connected line segments) on the left should result in one curve on the right, with a colored dot on that path corresponding to the color of the curve. The location of the dot will change as you hover in the plot on the right, allowing you to see which value in the curve corresponds to which location along the polyline.
Example 2: A time-varying mesh ¶
LineCrossSection
also allows working with time-varying meshes. Here we will modify the static TriMesh data with a time-varying random offset to demonstrate that the
LineCrossSection
also works for data that is evolving temporally.
sd_paths = [
[(-13037161., 3843454.), (-13041435., 3854275.), (-13045822., 3857996.),
(-13048664., 3857210.), (-13050429., 3855575.), (-13048385., 3849452.),
(-13048027., 3847896.)],
[(-13042975., 3850040.), (-13063919., 3844636.)]
]
filename = '../data/SanDiego_Mesh/SanDiego.3dm'
tris, verts = read_3dm_mesh(filename, skiprows=2)
points = gv.operation.project_points(gv.Points(verts, vdims=['z'], crs=ccrs.UTM(11)))
trimesh = gv.TriMesh((tris, points))
filename2 = '../data/SanDiego_Mesh/SanDiego_ovl.dat'
dfs = read_mesh2d(filename2)
points = gv.operation.project_points(gv.Points((verts.x, verts.y), crs=ccrs.UTM(11)))
def time_mesh(time):
depth_points = points.add_dimension('Velocity', 0, dfs[time].values[:, 0], vdim=True)
return gv.TriMesh((tris, depth_points), crs=ccrs.GOOGLE_MERCATOR)
time_dim = hv.Dimension('Time', values=sorted(dfs.keys()), default=3600)
meshes = hv.DynamicMap(time_mesh, kdims=time_dim)
sector2 = LineCrossSection(meshes, sd_paths, resolution=100)
sector2.view(cmap=cm_n.rainbow_r, shade=True)
You can use the scrubber controls to animate both plots, showing both the map-based data and the curve cross sections over time.
Example 3: Sampling a Surface ¶
Instead of sampling individual Curve elements for each time as in example 2, we can take cross-sections across time, returning an Image plotting the sampled value across both distance and time. The
SurfaceCrossSection
class is a simple subclass of
LineCrossSection
that overrides the
sample
method to apply the sampling for all time values and return an
Image
.
sector3 = SurfaceCrossSection(meshes, sd_paths[:1], resolution=100)
sector3.view(cmap=cm_n.rainbow_r, shade=False)
The scrubber will now only animate the plot on the left, as time has been laid out vertically on the plot on the right.
If you have a running Python process and want to draw your own line, first delete the existing one, because an overlay of non-transparent images will only show the one on top.